1. American Conference on Pharmacometics –San Diego, CA; April 3-6 2011 Population PK modeling incorporating enzyme induction mechanism to guide the design of optimal dose regimen for an antiviral drug Jing Yu, June Ke, Leo Kawai Jing Yu 1, June Ke 2, Leo Kawai 1 M&S - Novartis Pharma, Cambridge, MA; DMPK – Novartis Pharma, East Hanover, NJ 1 M&S - Novartis Pharma, Cambridge, MA; 2 DMPK – Novartis Pharma, East Hanover, NJ Objectives: Respiratory Syncytial Virus (RSV) infection is the leading cause of serious respiratory tract infections in infants, young children, elderly, and immuno-compromised patients throughout the world. RSV604 is a benzodiazepine analog that inhibits RSV replication in vitro. In vitro and clinical studies indicate that RSV604 systemic exposure had potential for a narrow therapeutic concentration range. Therefore, selection of optimal dose regimen required a model which precisely profiles individual PK in the therapeutic range. Methods: A linear three-compartment PK model with first-order absorption was built with Phase 1 study data including intravenous (IV; 10-200 mg) and oral (PO; 25-150 mg) doses, as population PK model with NONMEM. The linear model suggested a 450 mg loading dose followed by 150 mg BID to satisfy the therapeutic concentration range requirement. However, the highest dose group (200 mg) in the IV Phase 1 study and the 150 and 300 mg dose groups in a later conducted oral Phase 1 study showed the compound didn't accumulate after 5 days of dosing as projected from single dose data. As RSV604 was shown to moderately induce CYP3A4 by in vitro reporter gene assay, a clearance induction model, considering enzyme turn-over, was then examined, using in vitro enzyme induction parameters as initial estimations of in vivo model. Results: The enzyme induction model fitted all the data from Phase 1 studies well and eliminated the gaps observed in the initial linear model. This model predicted a 600 mg loading dose followed by 300 mg BID would rapidly reach the desired Cmin of 2000 ng/mL but decrease the potential safety risk. The prediction of this enzyme induction model was further validated by the PK data from subsequent clinical studies. Conclusions: Incorporation of an induction effect in the population PK model increased the accuracy in describing PK data from all Phase 1 studies and predicted an optimal dose regimen to be used in the next stage of drug development. Abstract Background Respiratory Syncytial Virus (RSV) infection is the leading cause of serious respiratory tract infections in infants and children, where bronchiolitis and pneumonia are the most common clinical syndromes. RSV infection is also a major cause of morbidity in immunocompromised patients, stem cell transplant patients, COPD patients, adult and children. These patients may experience high morbidity, increased mortality, and lasting sequelae from this infection. No vaccine exists and although aerolized ribavirin has been approved for use in infants with severe RSV infection, its effectiveness has been challenged and its well-known toxicity limits its utility. RSV604 is a benzodiazepine analogue and has potent inhibition against RSV viral replication, with in vitro IC 50 of 155 - 272 ng/mL RSV604 is 95% Protein bound, which leads to a ~12-fold shift in IC 50 at physiologic concentrations of albumin and AAG. CYP3A4 is a major responsible enzyme for the metabolism of RSV604; therefore, a potential 3A4 auto-induction might result in low compound exposure. Dose finding The target minimum exposure (Cmin) for in vivo activity based on the RSV604 IC 50 values and the protein binding was predicted to be ~2000 ng/mL. Previous clinical studies suggested RSV604 had the potential for a narrow therapeutic concentration range. Therefore a precise PK profile prediction will be critical for optimal dose regimen design. Using a linear PK model built with data from 2101 study, a 450 mg loading dose followed by 150 mg bid was predicted to satisfy the therapeutic concentration range requirement. Introduction Evidence of enzyme induction  2101 study: exposure at MD (AUC ss ) vs. SD (AUC inf ) tends to be low in high iv dose (200 mg); attenuated accumulation in higher dose  in vitro and animal data suggest CYP3A4 induction: in vitro EC 50 =~2750ng/mL  Linear model over-predicts data from the oral Phase 1 study (2103) cohorts 1 & 2 Enzyme auto-induction model  Concentration-dependent clearance: dCL int /dt = K reg *[1+ E max *C/(EC 50 +C) – CL int / Cl int,0 ], where C is concentration When C<< EC 50, it can be simplified to dCL int /dt = K reg *[1+ K e *C – CL int / Cl int,0 ], where K e = E max / EC 50  Estimated parameters K reg = 4.62 L/hr/hr; K e = 0.000427 /(ng/mL) Model building and validation  2103 study cohorts 1 & 2 data for model fitting (left) and cohort 3 data for model validation (right)  Estimated parameters Cl int,0 = 3.21 L/hr K reg = 4.62 L/hr/hr K e = 0.000427 /(ng/mL) New dose regimen for efficacy study  Potentially efficacious dose regimen suggested by induction model: 600 mg loading dose followed by 300 mg bid Conclusion The enzyme induction model represents the available clinical data well, although the real mechanism of the nonlinear PK profile needs to be further evaluated. With this model, we have reached a higher level of confidence in designing an optimal dose regimen, 600 mg loading dose followed by 300 mg bid, for the efficacy study in the next stage of drug development. Acknowledgements The whole RSV604 Team, especially Kathryn Bracken (PJM), Jens Praestgaard (CIS), Russ Wada (M&S), Daniel Stein and Tom Evans (TME).